INDEX

Explanations

expressions of deficiency or criticism

The neuron detects strongly negative or critical action words and qualifiers (e.g. “failing,” “not,” “more costly,” “oppose”) that signal criticism or problems.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 jajaja

-1.15

明日は

-1.05

さんへ

-1.05

Výhody

-1.03

iség

-1.02

 reportedly

-1.02

 those

-1.02

Ń

-1.01

 arguably

-1.01

いかがでしたか

-1.00

POSITIVE LOGITS

 слишком

1.44

1.41

 demasiado

1.40

too

1.36

“

1.27

 should

1.25

„

1.20

\"

1.17

«

1.16

 enough

1.16

Activations Density 0.047%